An Empirical Study on Task Documentation in Software Crowdsourcing on TopCoder Luis Vaz Igor Steinmacher Sabrina Marczak MunDDos Research Group School of Informatics, Computing, MunDDos Research Group School of Technology and Cyber Systems School of Technology PUCRS, Brazil Northern Arizona University, USA PUCRS, Brazil [email protected] [email protected] [email protected] Abstract—In the Software Crowdsourcingcompetitive model, challenged by the task description. Thus, in this context, crowd members seek for tasks in a platform and submit their the description (or documentation) of the task presented by solutions seeking rewards. In this model, the task description the platform becomes an important factor for the crowd is important to support the choice and the execution of a task. Despite its importance, little is known about the role of task members—who rely on it to choose and execute the tasks— description as support for these processes. To fill this gap, this and for the clients—who aim to receive complete and correct paper presents a study that explores the role of documentation on solutions to their problems. However, little is known about TopCoder platform, focusing on the task selection and execution. how the documentation influences the selection of the tasks We conducted a two-phased study with professionals that had and their subsequent development in the crowdsourcing model. no prior contact with TopCoder. Based on data collected with questionnaires, diaries, and a retrospective session, we could Seeking to contribute to filling this gap in the literature, this understand how people choose and perform the tasks, and the paper presents an empirical study aiming to understand the role of documentation in the platform. We could find that poorly role of documentation in the selection and development of specified or incomplete tasks lead developers to look for supple- tasks in software crowdsourcing. mentary material or invest more time and effort than initially To conduct this study, we focused on the TopCoder plat- estimated. To better support the crowd members, we proposed a model on how to structure the documentation that composes form [3], which implements the competitive model. We con- the task description in competitive software crowdsourcing. We ducted a two-phased research. In Phase 1 we conducted a evaluated the model with another set of professionals, again case study with professionals attending to a graduate course relying on questionnaires, reports, and a retrospective session. on Collaborative Software Development. The participants re- Results showed that although the documentation available cov- ceived an assignment that consisted of selecting, developing, ered the elements of the proposed model, the participants had issues to find the necessary information, suggesting the need for and submitting two tasks available on TopCoder. The data a reorganization. Participants agreed that the proposed model collection was carried out through questionnaires, diaries, would help them understand the task description. Therefore, our and a retrospective session. The results of Phase 1 showed study provides a better understanding of the importance of task that the quality of the task documentation, in particular its documentation in software crowdsourcing and points out what general description, influences the selection of the task by the information is important to the crowd. Index Terms—Software crowdsourcing, Task documentation, members of the crowd. Tasks that do not present clear and TopCoder, Empirical study objective description—without technology requirements, or environment setup instructions—demotivate developers, lead- I. INTRODUCTION ing them to abandon the task’s solution attempt. By analyzing the challenges and suggestions reported by the participants, we Software Crowdsourcing, defined as the act of outsourcing proposed a way to structure the information that composes the software development to an undefined and large group of task description in competitive software crowdsourcing. people—the crowd—through an open call, is a phenomenon The goal of Phase 2 was to evaluate the structure we that has been gaining attention [1], [2]. Crowdsourcing is proposed in Phase 1, in addition to gain more insights. To this mediated by a platform and is based on tasks that are executed end, we conducted another case study with professionals of by the crowd. The tasks can be distributed on-demand follow- a second session of the same course. We worked again with ing a recruitment model performed by the platform based on an assignment in which the 10 professionals had to choose the analysis of crowd members’ profiles; or in a competitive and submit a solution to a task in TopCoder. At this time, we model, in which the members register themselves for tasks and surveyed them about the challenges with documentation. We submit a solution aiming to receive a reward (e.g., financial). also asked them to write a report about their experience and In the competitive model, the crowd members can choose conducted a retrospective session as means to triangulate data. the tasks they will work on according to their own interest. For The results of Phase 2 indicate that although the information instance, they can filter the tasks according to the programming available in the tasks covered the elements of our proposed language, application domain, or simply because they feel model, the participants had issues to find the information, Fig. 2. Representation of task decomposition [7]. Fig. 1. Crowdsourcing model components. set of technical specifications followed by detailed instructions suggesting the need for a reorganization. Also, most of the on how to organize and submit the solution. participants agreed that the proposed structure would help The task represents the problem or part of the problem them identify and better understand different aspects of the defined by the client and is generally defined by the platform. task, reinforcing the need for improvement. This process of definition and decomposition into micro- Our results provide a better understanding of the importance tasks, illustrated in Figure 2, is considered one of the major of task documentation in software crowdsourcing and identify challenges of this development model [7]. This decomposition a set of information that may be provided to improve the process needs to safeguard that the tasks made available on experience of crowd members. Thus, this paper contributes the platform do not lose their characteristics and their interde- as follows: (i) listing the factors that influence the task choice pendence with the other parties that represent the problem [7]. and development; (ii) listing the quality characteristics of the When fragmenting the task into micro-tasks, the platform task documentation that are relevant to support the process of needs to ensure that each micro-task has sufficient information task choice and development; (iii) proposing a structured way to enable its development. The resulting task documentation to present the information when documenting crowdsourcing cannot be too specific that loses focus nor too broad that tasks; and (iv) evaluating how the description provided cur- challenges its comprehension. Tasks decomposition turns out rently adhere to the proposed structured model and how the to be a crucial factor for the tasks solution given that the structure would benefit the users. quality of the resulting documentation is likely to affect the crowd members’ performance [7]. II. BACKGROUND B. TopCoder Platform Howe [4] explains that the crowdsourcing phenomenon has Some crowdsourcing platforms are specialized in a certain its origins in the Open Source Software movement, evidencing software activity, like uTest1 and TestBirds2, which focus on that a motivated group with a common goal is able to create a testing; and others provide support the the entire development high-quality product. It involves outsourcing an activity so that cycle, such as GetACoder3 and TopCoder4. The latter stands a client reaches its business goals by leveraging a third-party out among the commercial platforms as the pioneer and for entity capable of solving a particular problem [4]. having thousands of active members [8]. With a similar idea behind it, Software Crowdsourcing is TopCoder was created in 2001 when its founder, Jack specialized in software development activities, and involves Hughes, proposed an innovative way of dealing with chal- the client, who proposes the task; the crowd, composed of lenges in recruiting talents and high turnover of profession- professionals willing to perform tasks; and the platform, which als [3]. Hughes aimed to reduce the costs and effort of his mediates the relationship between the other two elements [2], customers by reusing software components instead of building [5]. Among these elements is the task, which represents the a complete system from scratch. Thus, TopCoder defines activities proposed by the client, as illustrated in Figure 1. The a model that prioritizes the reuse of components to solve tasks are decomposed and managed by the platform, being customer problems and, when necessary, crowdsourced tasks carried out by the crowd [6]. to get solutions for the development of new components. As a way to enable this strategy, the platform becomes a mechanism A. Tasks in Software Crowdsourcing to attract new members and form a community of skilled A task in software crowdsourcing can represent a high programmers—the
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages10 Page
-
File Size-